Here is the flow of the additive attention function in terms of Neural Networks in order to give an intuition on how to implement it in Keras and Tensorflow